Google's Gemini Avatar lets you create an AI clone of yourself in minutes, sparking deepfake concerns

Google Rolls Out Gemini Avatar to Paid Subscribers

Google AI is expanding access to Gemini Avatar, a feature that allows users to create an AI clone of themselves for use in AI-generated videos 1

1

3

3

. First unveiled at Google I/O last month, the tool is now rolling out widely to paid Gemini subscribers, including those with AI Pro or AI Ultra accounts 4

4

. The feature captures not only physical likeness and voice but creates what early testers describe as unsettlingly realistic digital replicas. Users can access the tool through Google's Gemini app under Settings > Avatar, where a guided setup process takes just a few minutes to complete 3

3

.

How to Create an AI Clone of Myself Using Gemini

The process to create a personalized AI avatar starts with a straightforward enrollment procedure. Users aim their phone's camera at their face while following prompts to slowly move their head from side to side, then read a series of random numbers aloud 2

2

4

4

. The entire recording process takes only a couple of minutes, similar to enrolling in Face ID 2

2

. Once setup is complete, users can summon their avatar in Gemini chats by typing commands like @me or @[Your name], or by tapping the plus icon and selecting Avatar from the menu 1

1

3

3

. Video generation takes a couple of minutes, with users receiving a notification when their hyperrealistic deepfake video is ready 4

4

.

Powered by Multimodal Model Gemini Omni

The avatar system operates through the multimodal model Gemini Omni, Google's advanced AI that can accept images and videos as inputs and demonstrates far greater understanding of physics and the real world than previous models 2

2

. Users can create videos with varying levels of specificity, from simple prompts to detailed scenarios. One tester used the prompt "create a video of me wearing a t shirt with the Android Authority logo. Show me at the Google campus hanging out with the different Android figurines," producing remarkably accurate results 1

1

. The Videos page in Google's Gemini app includes several style templates such as "Anime," "Decades fashion," and "80's music video" to help users get started 1

1

.

Creepy Realism Raises Deepfake Content Concerns

Early users consistently describe the results as unsettling due to their realism. "My avatar's resemblance is honestly unsettling because it's so believable," one tester noted, adding that the facial movements and tone of voice are extremely realistic and could easily fool acquaintances 3

3

. Another user stated, "That looks and sounds an awful lot like me," though their daughter could identify it as AI-generated content 2

2

. The creepy accuracy extends beyond appearance to voice replication, with one tester observing that while the audio sounds authentic, it lacks natural cadence and personality, delivering content in monotone 4

4

. The potential for misuse has sparked immediate concerns about society's readiness for such accessible deepfake technology.

Safety Measures and Usage Restrictions

Google has implemented several safeguards recognizing the potential for misuse. Users must be at least 18 years old to create an avatar, and the account owner must be physically present during setup 3

3

. Every video generated includes Google's invisible SynthID watermark permanently embedded in the file, allowing verification of AI-generated content through Chrome or Google Search tools 3

3

4

4

. The feature is not available in the European Economic Area, Switzerland, or the United Kingdom, and currently only generates audio in English [4](https://lifehacker.com/tech/gemini-avatar-wider-rollout-3673898/]. There are also daily limits on video generation, with one AI Pro user being cut off after just three videos 2

2

. Users can delete their avatar at any time, and Google promises to wipe the selfie and voice data used to create it 2

2

. Videos are currently limited to 10 seconds with no editing capabilities, though Google is gathering feedback on potentially adding interactive editing features 4

4

.